version 7.0log using JVconenpcd_ci.log, replace#delimit ;*     ***************************************************************** *;*     ***************************************************************** *;*   File-Name:  JVconenpcd_ci.do                                        *;*   Date:       June 12, 2001                                           *;*   Author:     MRG                                                     *;*   Purpose:    Take Jackman and Volpert's 1996 BJPS piece and          *;*               calculate conditonal coefficients and standard errors   *;*   Input File: JVcorrectdata.dta                                       *;*   Output File:JVconenpcd_ci.log                                       *;*   Data Output:JVconenpcd_ci.dta                                       *;*   Previous file:                                                      *;*   Machine:    system7                                                 *;*     ****************************************************************  *;*     ****************************************************************  *;use h:\JackmanReplication\correctdata\JVcorrectdata.dta;summarize; *     ****************************************************************  *;*       Calculate conditional coefficients and standard errors for enp. *;*     ****************************************************************  *;set more off;tobit  lerps enp thresh threshenp unemp austria belgium denmark finland     france germany greece ireland italy netherlands norway portugal     spain sweden switzerland, ll(0) ltolerance (1e-5);*     ****************************************************************  *;*       thresh i.e. x2 = JH                                             *;*     ****************************************************************  *;set obs 360;generate JH=((_n-1)/1);    replace JH=. if _n>36;generate str1 txt="*";matrix b=e(b);matrix V=e(V);scalar b1=b[1,1];scalar b2=b[1,2];scalar b3=b[1,3];scalar varb1=V[1,1];scalar varb3=V[3,3];scalar covb1b3=V[1,3];set more off;scalar list b1 b2 b3 varb1 varb3 covb1b3;*     ****************************************************************  *;*       except for in left hand side of following, do search and replace*;*       for JH below                                                    *;*     ****************************************************************  *;gen conb=b1+b3*JH if _n<36;set more off;list conb in 1/20;*     ****************************************************************  *;*      These create conditional coefficients for b1 at JH, x3H#         *;*     ****************************************************************  *;*      We can write the following code as a template, then sub in the   *;*      names of the "variables" for JH *x3 .   "variables" because they *;*      won't be the real variables.  Instead, they'll be the variables  *;*      created to represent the relevant values of the modifying        *;*      variables.                                                       *;*     ****************************************************************  *;gen conse=sqrt(varb1+varb3*JH^2+2*covb1b3*JH)  if _n<360;set more off;list conse in 1/20;gen cont= conb/conse; list cont in 1/20; *     ****************************************************************  *;*      This creates a conditional t for each level of JHHH given which  *;*      x3H is chosen                                                    *;*     ****************************************************************  *;gen conbS = conb;*     ****************************************************************  *;*       These create "shadow" (thus the "S") variables for each of the  *;*       conditional b's.  Statistical significance indicated at 95%     *;*       level.                                                          *;*     ****************************************************************  *; replace conbS = . if abs(cont)<1.98;list conbS in 1/20;*     ****************************************************************  *;*      This creates missing values for all the insignificant shadow b's.*;*     ****************************************************************  *;*     ****************************************************************  *;*     ****************************************************************  *;*       Generate confidence intervals at the 95% level                  *;*     ****************************************************************  *;*     ****************************************************************  *;gen a=1.98*conse;gen top=conb+a;gen bottom=conb-a;set textsize 100;label var JH "Effective Threshold";label var conb "Coefficient for ENPP";label var conbS "* indicates p<0.05";*     ****************************************************************  *;*       Graph the effect of ENPP on ERPS conditional on THRESH           *;*     ****************************************************************  *;gprefs set window scheme blackbg;gprefs set window update;gprefs set custom1 pen2_thick 9;graph  conb top bottom JH, psize(100) noaxis ylab( -3 -2  -1  0  1 2)     xlab(0 5 10 15 20 25 30 35) yline(0) pen(955) symbol(...) c(l l[_.] l[_.])sort ; set textsize 100;*     ****************************************************************  *;*       Note that in the graph there is a large area where the          *;*       coefficient on ENP is not statistically significant.  This      *;*       range is from 0 to 31.7 on the effective threshold scale.       *;*       92 observations fall in this range which is over 90% of the     *;*       observations. This means that JV's conclusion about the impact  *;*       of ENP is not accurate.                                         *;*     ****************************************************************  *;save    JVconenpcd_ci, replace;exit;